wavelan_cs.p.h

linux-2.6.15.6

/*
 *	Wavelan Pcmcia driver
 *
 *		Jean II - HPLB '96
 *
 * Reorganisation and extension of the driver.
 *
 * This file contain all definition and declarations necessary for the
 * wavelan pcmcia driver. This file is a private header, so it should
 * be included only on wavelan_cs.c !!!
 */

#ifndef WAVELAN_CS_P_H
#define WAVELAN_CS_P_H

/************************** DOCUMENTATION **************************/
/*
 * This driver provide a Linux interface to the Wavelan Pcmcia hardware
 * The Wavelan is a product of Lucent (http://www.wavelan.com/).
 * This division was formerly part of NCR and then AT&T.
 * Wavelan are also distributed by DEC (RoamAbout DS)...
 *
 * To know how to use this driver, read the PCMCIA HOWTO.
 * If you want to exploit the many other fonctionalities, look comments
 * in the code...
 *
 * This driver is the result of the effort of many peoples (see below).
 */

/* ------------------------ SPECIFIC NOTES ------------------------ */
/*
 * Web page
 * --------
 *	I try to maintain a web page with the Wireless LAN Howto at :
 *	    http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/Wavelan.html
 *
 * SMP
 * ---
 *	We now are SMP compliant (I eventually fixed the remaining bugs).
 *	The driver has been tested on a dual P6-150 and survived my usual
 *	set of torture tests.
 *	Anyway, I spent enough time chasing interrupt re-entrancy during
 *	errors or reconfigure, and I designed the locked/unlocked sections
 *	of the driver with great care, and with the recent addition of
 *	the spinlock (thanks to the new API), we should be quite close to
 *	the truth.
 *	The SMP/IRQ locking is quite coarse and conservative (i.e. not fast),
 *	but better safe than sorry (especially at 2 Mb/s ;-).
 *
 *	I have also looked into disabling only our interrupt on the card
 *	(via HACR) instead of all interrupts in the processor (via cli),
 *	so that other driver are not impacted, and it look like it's
 *	possible, but it's very tricky to do right (full of races). As
 *	the gain would be mostly for SMP systems, it can wait...
 *
 * Debugging and options
 * ---------------------
 *	You will find below a set of '#define" allowing a very fine control
 *	on the driver behaviour and the debug messages printed.
 *	The main options are :
 *	o WAVELAN_ROAMING, for the experimental roaming support.
 *	o SET_PSA_CRC, to have your card correctly recognised by
 *	  an access point and the Point-to-Point diagnostic tool.
 *	o USE_PSA_CONFIG, to read configuration from the PSA (EEprom)
 *	  (otherwise we always start afresh with some defaults)
 *
 * wavelan_cs.o is darn too big
 * -------------------------
 *	That's true ! There is a very simple way to reduce the driver
 *	object by 33% (yes !). Comment out the following line :
 *		#include <linux/wireless.h>
 *	Other compile options can also reduce the size of it...
 *
 * MAC address and hardware detection :
 * ----------------------------------
 *	The detection code of the wavelan chech that the first 3
 *	octets of the MAC address fit the company code. This type of
 *	detection work well for AT&T cards (because the AT&T code is
 *	hardcoded in wavelan_cs.h), but of course will fail for other
 *	manufacturer.
 *
 *	If you are sure that your card is derived from the wavelan,
 *	here is the way to configure it :
 *	1) Get your MAC address
 *		a) With your card utilities (wfreqsel, instconf, ...)
 *		b) With the driver :
 *			o compile the kernel with DEBUG_CONFIG_INFO enabled
 *			o Boot and look the card messages
 *	2) Set your MAC code (3 octets) in MAC_ADDRESSES[][3] (wavelan_cs.h)
 *	3) Compile & verify
 *	4) Send me the MAC code - I will include it in the next version...
 *
 */

/* --------------------- WIRELESS EXTENSIONS --------------------- */
/*
 * This driver is the first one to support "wireless extensions".
 * This set of extensions provide you some way to control the wireless
 * caracteristics of the hardware in a standard way and support for
 * applications for taking advantage of it (like Mobile IP).
 *
 * You will need to enable the CONFIG_NET_RADIO define in the kernel
 * configuration to enable the wireless extensions (this is the one
 * giving access to the radio network device choice).
 *
 * It might also be a good idea as well to fetch the wireless tools to
 * configure the device and play a bit.
 */

/* ---------------------------- FILES ---------------------------- */
/*
 * wavelan_cs.c :	The actual code for the driver - C functions
 *
 * wavelan_cs.p.h :	Private header : local types / vars for the driver
 *
 * wavelan_cs.h :	Description of the hardware interface & structs
 *
 * i82593.h :		Description if the Ethernet controller
 */

/* --------------------------- HISTORY --------------------------- */
/*
 * The history of the Wavelan drivers is as complicated as history of
 * the Wavelan itself (NCR -> AT&T -> Lucent).
 *
 * All started with Anders Klemets <[e-mail]>,
 * writting a Wavelan ISA driver for the MACH microkernel. Girish
 * Welling <[e-mail]> had also worked on it.
 * Keith Moore modify this for the Pcmcia hardware.
 * 
 * Robert Morris <[e-mail]> port these two drivers to BSDI
 * and add specific Pcmcia support (there is currently no equivalent
 * of the PCMCIA package under BSD...).
 *
 * Jim Binkley <[e-mail]> port both BSDI drivers to FreeBSD.
 *
 * Bruce Janson <[e-mail]> port the BSDI ISA driver to Linux.
 *
 * Anthony D. Joseph <[e-mail]> started modify Bruce driver
 * (with help of the BSDI PCMCIA driver) for PCMCIA.
 * Yunzhou Li <[e-mail]> finished is work.
 * Joe Finney <[e-mail]> patched the driver to start
 * correctly 2.00 cards (2.4 GHz with frequency selection).
 * David Hinds <[e-mail]> integrated the whole in his
 * Pcmcia package (+ bug corrections).
 *
 * I (Jean Tourrilhes - [e-mail]) then started to make some
 * patchs to the Pcmcia driver. After, I added code in the ISA driver
 * for Wireless Extensions and full support of frequency selection
 * cards. Now, I'm doing the same to the Pcmcia driver + some
 * reorganisation.
 * Loeke Brederveld <[e-mail]> from Lucent has given me
 * much needed informations on the Wavelan hardware.
 */

/* By the way : for the copyright & legal stuff :
 * Almost everybody wrote code under GNU or BSD license (or alike),
 * and want that their original copyright remain somewhere in the
 * code (for myself, I go with the GPL).
 * Nobody want to take responsibility for anything, except the fame...
 */

/* --------------------------- CREDITS --------------------------- */
/*
 * Credits:
 *    Special thanks to Jan Hoogendoorn of AT&T GIS Utrecht and
 *	Loeke Brederveld of Lucent for providing extremely useful
 *	information about WaveLAN PCMCIA hardware
 *
 *    This driver is based upon several other drivers, in particular:
 *	David Hinds' Linux driver for the PCMCIA 3c589 ethernet adapter
 *	Bruce Janson's Linux driver for the AT-bus WaveLAN adapter
 *	Anders Klemets' PCMCIA WaveLAN adapter driver
 *	Robert Morris' BSDI driver for the PCMCIA WaveLAN adapter
 *
 * Additional Credits:
 *
 *    This software was originally developed under Linux 1.2.3
 *	(Slackware 2.0 distribution).
 *    And then under Linux 2.0.x (Debian 1.1 -> 2.2 - pcmcia 2.8.18+)
 *	with an HP OmniBook 4000 and then a 5500.
 *
 *    It is based on other device drivers and information either written
 *    or supplied by:
 *	James Ashton ([e-mail]),
 *	Ajay Bakre ([e-mail]),
 *	Donald Becker ([e-mail]),
 *	Jim Binkley <[e-mail]>,
 *	Loeke Brederveld <[e-mail]>,
 *	Allan Creighton ([e-mail]),
 *	Brent Elphick <[e-mail]>,
 *	Joe Finney <[e-mail]>,
 *	Matthew Geier ([e-mail]),
 *	Remo di Giovanni ([e-mail]),
 *	Mark Hagan ([e-mail]),
 *	David Hinds <[e-mail]>,
 *	Jan Hoogendoorn (c/o [e-mail]),
 *      Bruce Janson <[e-mail]>,
 *	Anthony D. Joseph <[e-mail]>,
 *	Anders Klemets ([e-mail]),
 *	Yunzhou Li <[e-mail]>,
 *	Marc Meertens ([e-mail]),
 *	Keith Moore,
 *	Robert Morris ([e-mail]),
 *	Ian Parkin ([e-mail]),
 *	John Rosenberg ([e-mail]),
 *	George Rossi ([e-mail]),
 *	Arthur Scott ([e-mail]),
 *	Stanislav Sinyagin <[e-mail]>
 *	Peter Storey,
 *	Jean Tourrilhes <[e-mail]>,
 *	Girish Welling ([e-mail])
 *	Clark Woodworth <[e-mail]>
 *	Yongguang Zhang <[e-mail]>...
 */

/* ------------------------- IMPROVEMENTS ------------------------- */
/*
 * I proudly present :
 *
 * Changes made in 2.8.22 :
 * ----------------------
 *	- improved wv_set_multicast_list
 *	- catch spurious interrupt
 *	- correct release of the device
 *
 * Changes mades in release :
 * ------------------------
 *	- Reorganisation of the code, function name change
 *	- Creation of private header (wavelan_cs.h)
 *	- Reorganised debug messages
 *	- More comments, history, ...
 *	- Configure earlier (in "insert" instead of "open")
 *        and do things only once
 *	- mmc_init : configure the PSA if not done
 *	- mmc_init : 2.00 detection better code for 2.00 init
 *	- better info at startup
 *	- Correct a HUGE bug (volatile & uncalibrated busy loop)
 *	  in wv_82593_cmd => config speedup
 *	- Stop receiving & power down on close (and power up on open)
 *	  use "ifconfig down" & "ifconfig up ; route add -net ..."
 *	- Send packets : add watchdog instead of pooling
 *	- Receive : check frame wrap around & try to recover some frames
 *	- wavelan_set_multicast_list : avoid reset
 *	- add wireless extensions (ioctl & get_wireless_stats)
 *	  get/set nwid/frequency on fly, info for /proc/net/wireless
 *	- Suppress useless stuff from lp (net_local), but add link
 *	- More inlines
 *	- Lot of others minor details & cleanups
 *
 * Changes made in second release :
 * ------------------------------
 *	- Optimise wv_85893_reconfig stuff, fix potential problems
 *	- Change error values for ioctl
 *	- Non blocking wv_ru_stop() + call wv_reset() in case of problems
 *	- Remove development printk from wavelan_watchdog()
 *	- Remove of the watchdog to wavelan_close instead of wavelan_release
 *	  fix potential problems...
 *	- Start debugging suspend stuff (but it's still a bit weird)
 *	- Debug & optimize dump header/packet in Rx & Tx (debug)
 *	- Use "readb" and "writeb" to be kernel 2.1 compliant
 *	- Better handling of bogus interrupts
 *	- Wireless extension : SETSPY and GETSPY
 *	- Remove old stuff (stats - for those needing it, just ask me...)
 *	- Make wireless extensions optional
 *
 * Changes made in third release :
 * -----------------------------
 *	- cleanups & typos
 *	- modif wireless ext (spy -> only one pointer)
 *	- new private ioctl to set/get quality & level threshold
 *	- Init : correct default value of level threshold for pcmcia
 *	- kill watchdog in hw_reset
 *	- more 2.1 support (copy_to/from_user instead of memcpy_to/fromfs)
 *	- Add message level (debug stuff in /var/adm/debug & errors not
 *	  displayed at console and still in /var/adm/messages)
 *
 * Changes made in fourth release :
 * ------------------------------
 *	- multicast support (yes !) thanks to Yongguang Zhang.
 *
 * Changes made in fifth release (2.9.0) :
 * -------------------------------------
 *	- Revisited multicast code (it was mostly wrong).
 *	- protect code in wv_82593_reconfig with dev->tbusy (oups !)
 *
 * Changes made in sixth release (2.9.1a) :
 * --------------------------------------
 *	- Change the detection code for multi manufacturer code support
 *	- Correct bug (hang kernel) in init when we were "rejecting" a card 
 *
 * Changes made in seventh release (2.9.1b) :
 * ----------------------------------------
 *	- Update to wireless extensions changes
 *	- Silly bug in card initial configuration (psa_conf_status)
 *
 * Changes made in eigth release :
 * -----------------------------
 *	- Small bug in debug code (probably not the last one...)
 *	- 1.2.13 support (thanks to Clark Woodworth)
 *
 * Changes made for release in 2.9.2b :
 * ----------------------------------
 *	- Level threshold is now a standard wireless extension (version 4 !)
 *	- modules parameters types for kernel > 2.1.17
 *	- updated man page
 *	- Others cleanup from David Hinds
 *
 * Changes made for release in 2.9.5 :
 * ---------------------------------
 *	- byte count stats (courtesy of David Hinds)
 *	- Remove dev_tint stuff (courtesy of David Hinds)
 *	- Others cleanup from David Hinds
 *	- Encryption setting from Brent Elphick (thanks a lot !)
 *	- 'base' to 'u_long' for the Alpha (thanks to Stanislav Sinyagin)
 *
 * Changes made for release in 2.9.6 :
 * ---------------------------------
 *	- fix bug : no longuer disable watchdog in case of bogus interrupt
 *	- increase timeout in config code for picky hardware
 *	- mask unused bits in status (Wireless Extensions)
 *
 * Changes integrated by Justin Seger <[e-mail]> & David Hinds :
 * -----------------------------------------------------------------
 *	- Roaming "hack" from Joe Finney <[e-mail]>
 *	- PSA CRC code from Bob Gray <[e-mail]>
 *	- Better initialisation of the i82593 controller
 *	  from Joseph K. O'Sullivan <[e-mail]>
 *
 * Changes made for release in 3.0.10 :
 * ----------------------------------
 *	- Fix eject "hang" of the driver under 2.2.X :
 *		o create wv_flush_stale_links()
 *		o Rename wavelan_release to wv_pcmcia_release & move up
 *		o move unregister_netdev to wavelan_detach()
 *		o wavelan_release() no longer call wavelan_detach()
 *		o Suppress "release" timer
 *		o Other cleanups & fixes
 *	- New MAC address in the probe
 *	- Reorg PSA_CRC code (endian neutral & cleaner)
 *	- Correct initialisation of the i82593 from Lucent manual
 *	- Put back the watchdog, with larger timeout
 *	- TRANSMIT_NO_CRC is a "normal" error, so recover from it
 *	  from Derrick J Brashear <[e-mail]>
 *	- Better handling of TX and RX normal failure conditions
 *	- #ifdef out all the roaming code
 *	- Add ESSID & "AP current address" ioctl stubs
 *	- General cleanup of the code
 *
 * Changes made for release in 3.0.13 :
 * ----------------------------------
 *	- Re-enable compilation of roaming code by default, but with
 *	  do_roaming = 0
 *	- Nuke `nwid=nwid^ntohs(beacon->domain_id)' in wl_roam_gather
 *	  at the demand of John Carol Langford <[e-mail]>
 *	- Introduced WAVELAN_ROAMING_EXT for incomplete ESSID stuff.
 *
 * Changes made for release in 3.0.15 :
 * ----------------------------------
 *	- Change e-mail and web page addresses
 *	- Watchdog timer is now correctly expressed in HZ, not in jiffies
 *	- Add channel number to the list of frequencies in range
 *	- Add the (short) list of bit-rates in range
 *	- Developp a new sensitivity... (sens.value & sens.fixed)
 *
 * Changes made for release in 3.1.2 :
 * ---------------------------------
 *	- Fix check for root permission (break instead of exit)
 *	- New nwid & encoding setting (Wireless Extension 9)
 *
 * Changes made for release in 3.1.12 :
 * ----------------------------------
 *	- reworked wv_82593_cmd to avoid using the IRQ handler and doing
 *	  ugly things with interrupts.
 *	- Add IRQ protection in 82593_config/ru_start/ru_stop/watchdog
 *	- Update to new network API (softnet - 2.3.43) :
 *		o replace dev->tbusy (David + me)
 *		o replace dev->tstart (David + me)
 *		o remove dev->interrupt (David)
 *		o add SMP locking via spinlock in splxx (me)
 *		o add spinlock in interrupt handler (me)
 *		o use kernel watchdog instead of ours (me)
 *		o verify that all the changes make sense and work (me)
 *	- Re-sync kernel/pcmcia versions (not much actually)
 *	- A few other cleanups (David & me)...
 *
 * Changes made for release in 3.1.22 :
 * ----------------------------------
 *	- Check that SMP works, remove annoying log message
 *
 * Changes made for release in 3.1.24 :
 * ----------------------------------
 *	- Fix unfrequent card lockup when watchdog was reseting the hardware :
 *		o control first busy loop in wv_82593_cmd()
 *		o Extend spinlock protection in wv_hw_config()
 *